The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were identified. The skill follows standard software testing practices, including build pinning, evidence collection, and safe data handling.- [PROMPT_INJECTION]: The skill processes untrusted data from defect reports and status ledgers stored on the file system, which constitutes a surface for indirect prompt injection.
Ingestion points: bugs/ directory, report.md files, and consolidated ledger files.
Boundary markers: The skill designates the "Minimal repro steps" section as the authoritative source for the agent's actions, creating a logical boundary.
Capability inventory: The skill performs file system writes to update reports/ledgers and interacts with external tools such as Jira.
Sanitization: The workflow relies on manual cross-referencing of evidence and build pinning rather than automated input sanitization.

fix-verifier